Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.512619
Title: Feature-based object tracking in maritime scenes
Author: Voles, P.
Awarding Body: Bournemouth University
Current Institution: Bournemouth University
Date of Award: 2005
Availability of Full Text:
Access through EThOS:
Access through Institution:
Abstract:
A monitoring of presence, location and activity of various objects on the sea is essential for maritime navigation and collision avoidance. Mariners normally rely on two complementary methods of the monitoring: radar and satellite-based aids and human observation. Though radar aids are relatively accurate at long distances, their capability of detecting small, unmanned or non-metallic craft that generally do not reflect radar waves sufficiently enough, is limited. The mariners, therefore, rely in such cases on visual observations. The visual observation is often facilitated by using cameras overlooking the sea that can also provide intensified infra-red images. These systems or nevertheless merely enhance the image and the burden of the tedious and error-prone monitoring task still rests with the operator. This thesis addresses the drawbacks of both methods by presenting a framework consisting of a set of machine vision algorithms that facilitate the monitoring tasks in maritime environment. The framework detects and tracks objects in a sequence of images captured by a camera mounted either on a board of a vessel or on a static platform over-looking the sea. The detection of objects is independent of their appearance and conditions such as weather and time of the day. The output of the framework consists of locations and motions of all detected objects with respect to a fixed point in the scene. All values are estimated in real-world units, i. e. location is expressed in metres and velocity in knots. The consistency of the estimates is maintained by compensating for spurious effects such as vibration of the camera. In addition, the framework continuously checks for predefined events such as collision threats or area intrusions, raising an alarm when any such event occurs. The development and evaluation of the framework is based on sequences captured under conditions corresponding to a designated application. The independence of the detection and tracking on the appearance of the sceneand objects is confirmed by a final cross-validation of the framework on previously unused sequences. Potential applications of the framework in various areas of maritime environment including navigation, security, surveillance and others are outlined. Limitations to the presented framework are identified and possible solutions suggested. The thesis concludes with suggestions to further directions of the research presented.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.512619  DOI: Not available
Keywords: General Engineering
Share: